Data processing limits?

RE: Data processing limits?

Audun — Mon, 18 Nov 2013 14:17:26 GMT

According to the Cortex-M0 specs (www.arm.com/.../cortex-m0.php) you should see up to 1.21 DMIPS / MHz. The nRF51 CPU clock is 16 MHz.

If you want to do a more detailed analysis or test case, you can use gcc-arm or the free eval version of Keil (www.keil.com/.../mdk.asp) to compile some examples and see the code size. If there are very specific procedures you have in mind, it's possible to use the Keil simulator (or debug an nRF51) and step through the assembly. There is no memory cache in the nRF51, and most instructions are single-cycle (infocenter.arm.com/.../index.jsp, so the number of instructions is a strong indicator of performance.

Especially when dealing with floating point data it is useful to see the assembly or linker output to get a sense of complexity and performance.

RE: Data processing limits?

Elena — Mon, 18 Nov 2013 12:58:23 GMT

Hello,

Thanks for your quick response, I was trying to figure, the complexity of processing I can reach with that kind of processor,

RE: Data processing limits?

Audun — Mon, 18 Nov 2013 12:40:33 GMT

Hi,

apart from speed, I suppose the main issue would be code space and RAM. nRF51 have 256 kByte flash and 16 kByte RAM in total. If you are using Bluetooth, the stack will reserve 80 kByte of flash and 8 kByte of RAM.

I'm not exactly sure what you mean by data processing limits. The Cortex M0 is a 32-bit architecture. It does not have FPU, so all floating point operations are done in software, which in effect means that floating point operations will be very much slower than integer operations. The instruction set does not have DSP-type instructions, but that impacts speed, not functionality.